Adan Vielma
Co-founder & CTO, Screened · Singapore
What it takes to make AI conversations work reliably in production.

At Screened, Adan builds AI workflows that connect candidate interviews with the evidence recruiters use to make decisions. He previously worked at Singapore’s Open Government Products and delivered a workshop for its Build for Good accelerator.
Speaking topics
For engineers, technical founders and teams building AI products. Technical talks, demonstrations, panels and workshops.
The interview finished. The system said it didn’t.
A candidate finishes an AI interview, but the system marks it incomplete. Adan unpacks a production failure to show how to test completion, handle repeat attempts and recover from saved evidence without relying on the AI to say a particular closing phrase.
The candidate corrected themselves. Did the voice agent correct the record?
A candidate says they “managed ten people,” then clarifies that they “coordinated ten people.” Using this fictional example, Adan shows how to check whether the correction carries through the voice conversation, transcript and final assessment.